Every number here is a probability, not a call. Over the last 14 days the next-pitch model was graded on 34,173 real pitches: the top pick was right 39% of the time, and stated probabilities landed within 1.4 points of how often things actually happened. It grades itself every night and corrects what it can.
Win probability
Baltimore Orioles 64%, give or take 0.6 points of simulation noise · anchored to 115,696 real games in this exact situation, then adjusted for these two teams.
Projected score
8.6 – 5.6expected runs
3–14Baltimore Orioles, 80% range
2–10Colorado Rockies, 80% range
14.1total runs
Most likely finals 6–7, 5–4, 6–5 (each under 2%, which is the point: no single score is likely). From 6,000 simulated games.
Starting pitchers
Trevor RogersBaltimore Orioles
Innings5.0 3.3–6.7
Strikeouts5.3 3–8
Walks1.5 0–3
Hits allowed6.3 3–10
Home runs0.7 0–2
Runs allowed3.0 0–6
Pitches87 63–110
Tomoyuki SuganoColorado Rockies
Innings4.5 3.0–6.0
Strikeouts3.7 1–6
Walks1.5 0–3
Hits allowed7.2 4–11
Home runs1.4 0–3
Runs allowed4.3 1–8
Pitches85 63–107
Each figure is the median of the simulated games with an 80% range beside it. Pitch counts are estimated from batters faced (fitted on 17,200 real outings; typical error 8 pitches). These assume the starter is not pulled early for a reason the model cannot see — an injury, a blowout, or a quick hook.
Park and weather
1.28×park
1.28×run environment
the park favours hitters (1.28x runs). Park factors are measured from 25,110 games by comparing each club at home with the same club on the road, so a good offence does not inflate its own park. Player rates are park-neutralised before this is applied, or the park would count twice.
What this does not know
- starters are pulled around their own typical workload and a bullpen composite finishes, but individual relievers are not modelled, so a specific closer in a specific inning is not
- defence beyond what is already in the pitcher's line is not modelled, and neither are pinch hitters or intentional walks as a choice
- lineup not posted; using the order most recently used
- bullpen ordered by measured availability: Chase Dollander (long, first appearance in a while, 45% likely); Blas Castaño (long, first appearance in a while, 45% likely); Zach Agnos (long, 4d rest, 45% likely); Tanner Gordon (long, 2d rest, 40% likely)
- the catcher loses 3.5 strikes per 100 borderline pitches
- Baltimore Orioles defence lets through 0.3 plays per 100 balls in play
- Colorado Rockies defence lets through 0.7 plays per 100 balls in play
- run environment: the park favours hitters (1.28x runs)
- no weather posted yet; temperature and wind are not in this projection
Full game page · state as of 20260902_013941